Mutations in a newly identified Drosophila melanogaster gene, mago nashi, disrupt germ cell formation and result in the formation of mirror-image symmetrical double abdomen embryos

With the possible exception of mammalian eggs, pattern formation in the early embryo is initiated by maternally encoded cytoplasmic factors (determinants) asymmetrically distributed in the egg. By providing positional information, cytoplasmic determinants re- strict early embryonic cells to particular developmental pathways necessary for the establishment of the major body axes of the organism and, in Drosophila and Xenopus, cytoplasmic factors localized within the germ plasm have been implicated in the determination of the germ cell lineage (Wilson, 1925; Davidson, 1986; Wilkins, 1986). Presumably, these determinants specify cell fates by directly or indirectly regulating the expression of genes within cells that inherit them. Thus, early spatial differences in the pattern of cellular differentiation can reflect: (1) localization of cytoplas- mic determinants during oogenesis (Wilson, 1925; Boswell and Mahowald, 1985a; Davidson, 1986); (2) reorganization of the cytoplasmic constituents after fertilization resulting in local activation and/or localiz- ation of these determinants (Davidson, 1986; Klingler et al. 1988; Sprenger et al. 1989; Strecker et al. 1989); and (3) a graded distribution of these factors (Davidson, 1986; Driever and Nüsslein-Volhard, 1988a,b;Driever, et al. 1990). In any case, it is the underlying spatial organization of these determinants within the egg that will ultimately direct early embryonic events. Thus, how the spatial organization of the egg is established and how this organization becomes reflected in the final pattern of the embryo remains a fundamental problem in developmental biology.

In Drosophila, the posterior pole plasm has been demonstrated to be essential both for germ cell determination and delineation of the anteroposterior axis of the insect embryo (Illmensee and Mahowald, 1974, 1976; Frohnhöfer et al. 1986). Results of constriction and cytoplasm removal and/or transfer indicate that two morphogenetic centers, one at each pole of the embryo, interact to establish the longitudi- nal body pattern (Schubiger et al. 1977; Schubiger and Newman, 1982; Frohnhöfer et al. 1986; Sugiyama and Okada, 1990). Removal of cytoplasm from the anterior pole results in the reduction or complete loss of anterior structures and a replacement of these structures by asegmental tail (telson) structures. Moreover, when posterior pole cytoplasm is transplanted to the anterior pole of an embryo from which anterior cytoplasm has been removed double abdomen embryos are formed. In contrast, removal of cytoplasm from the posterior tip results in a loss of abdominal segments, but is not accompanied by the acquisition of anterior structures at this tip or a reduction in telson structures. Thus, cytoplasm removal and/or transfer experiments indi- cate that localized anterior and posterior factors exhibit long-range organizing abilities with respect to longitudi- nal axis formation in the embryo. Furthermore, these factors have the properties expected for morphogens in that they exhibit inductive properties when trans- planted to ectopic sites within the embryo.

Genetic analysis in Drosophila melanogaster has defined more than 20 strict maternal effect genes that are required for anteroposterior axis formation (Akam, 1987; Nüsslein-Volhard et al. 1987; Ingham, 1988; Manseau and Schüpbach, 1989a). These genes are divided into categories based on the regions of the normal embryonic pattern elements altered in embryos from mutant mothers, and can be phenocopied by cytoplasm removal and/or transfer experiments. For example, anterior group genes such as bicoid (bed) are required for the formation of head and thoracic segments (Frohnhöfer and Nüsslein-Volhard, 1986). Thus, embryos from homozygous bcd^E1 mothers lack head and thoracic segments, which are replaced by telson structures. Genes within the posterior group (nanos and pumilio) are necessary for segmentation of the embryonic abdomen (Lehmann and Nüsslein- Volhard, 1987; Nüsslein-Volhard et al. 1987) and a subset (cappuccino, oskar, staufen, spire, tudor, vasa and valais) are required both for abdominal segmen- tation and pole cell formation (Boswell and Mahowald, 1985b; Lehmann and Nüsslein-Volhard, 1986; Schüp- bach and Wieschaus, 1986; Manseau and Schüpbach, 1989b). Mutations in the spatial coordinates genes (bicaudal, Bicaudal-C and Bicaudal-D) result in the replacement of head and thoracic segments by telson and abdominal segments in reverse polarity to the normal posterior structures (Bull, 1966; Nüsslein- Volhard, 1977; Mohler and Wieschaus, 1986). A class that has not been phenocopied is the terminal group genes (e.g., torso and trunk) in which the terminal structures (asegmental head and telson) are deleted (Schüpbach and Wieschaus, 1986; Nüsslein-Volhard et al. 1987). Except for the terminal group genes, the observed pattern defects are similar to those obtained in experimental manipulations in which the anterior and posterior organizing centers are disrupted (Frohn- höfer et al. 1986; Nüsslein-Volhard et al. 1987). Therefore, the genes within the posterior and anterior groups are currently thought to be required for the establishment and/or function of their respective organizing centers.

In a continued effort to identify and characterize genes essential for pattern formation and germ cell determination in Drosophila melanogaster, we have isolated maternal effect mutations that disrupt these processes. Here we describe the genetic and develop- mental analysis of a previously unidentified strict maternal effect locus, mago nashi (mago). We provide evidence that mutations of the locus disrupt an oogenetic process required for normal functioning of the germ plasm and that these same mutations result in the production of a novel gene function capable of altering the fate of cells along the entire anteroposterior axis of the embryo. The phenotypic analysis of inviable embryos from mutant mago mothers indicates that the mago locus is unique among maternal effect genes required for the establishment of the anteroposterior axis, both in the variety and types of segmentation pattern abnormalities that it produces when mutated.

Drosophila melanogaster strains were cultured on standard cornmeal, agar, molasses and yeast medium in 180ml plastic bottles or 8 dram vials. Except for the mago nashi (mago) alleles and where otherwise indicated, all mutations and balancers are described in Lindsley and Grell (1968). The original mago mutation, mago¹, was N-ethyl-N-nitrosourea (ENU)-induced in a cn bw sp background, as described by Lee (1987). mago¹ was identified in a genetic screen in which females homozygous for the mutagenized cn bw sp chromo- some produced sterile F₁ progeny. The mating scheme used was similar to that described below for the isolation of mago² and mago³. Approximately 2600 second chromosomes were screened and we obtained one allele of mago and two alleles of tud (tud was used to monitor the efficacy of the screen). The name mago nashi means without grandchildren, in the Japanese language.

The mating scheme used to identify mago² and mago³ is described below. 100 males of the genotype cn bw sp/cn bw sp or Sco cn bw sp balanced by In(2LR)SM6a, al²Cy dp^lvlpr cn^2Psp² (SM6a) were fed 25 DIM ethyl methanesulfonate (EMS) in a 1 % sucrose solution for 24 h (Lewis and Bacher, 1968). In new bottles containing standard medium, 10 mutagenized males of the genotype cn bw sp (or Sco cn bw sp)/SM6a were crossed to 20 virgin females of the genotype nub b Sco lt stw³/SM6a. The males were discarded after four days. F₁cn bw sp (or Sco cn bw sp)/SM6a males were collected and crossed individually to virgin cv-2 mago¹bw sp/SM6a females in 8 dram vials. F₂cn bw sp (or Sco cn bw sp)/SM6a males were collected and later used to construct stocks. F₂ females of the genotype cn bw sp (or Sco cn bw sp)/cv-2 mago¹bw sp were selected and mated to sibling males for 4 days, after which all adults were discarded. The Fj progeny of cn bw sp (or Seo cn bw sp)/cv-2 mago¹bw sp females were transferred to new vials and, after 14 days, the vials were inspected for the presence of eggs. When no eggs were found, the reproductive organs of the males and females were inspected by light microscopy. In this manner, we identified mago alleles fully penetrant for the sterility phenotype.

To obtain a precise localization of the locus, we determined its cytological position by testing the ability of various chromosomal deletions to complement mago¹. We initially localized the locus to the polytene chromosome interval 57B5;58Bl–2, deleted by Df(2R)D17 (O’Donnell et al. 1989). Of the available chromosomal deletions within this polytene chromosome interval, mago fails to complement Df(2R)PC18 (57B16;57C6), Df(2R)F36 (57B17–20;57C6), and Df(2R)PL3 (57B20;57Dl,2). but complements Df(2R)CC2 (57C2;57C5) and Df(2R)MPl (57C2;57D1,2) (O’Donnell et al. 1989); therefore, the mago locus must reside in the polytene chromosome interval 57B20;57C2. This localization was further confirmed by the isolation of recombinants between crossveinless on chromosome 2 (cv-2) and mago¹, and Punch (Pu) and mago¹. The cv-2 and Pu loci map at positions 96.2 and 97, respectively. Thus, both recombination and deletion mapping place the locus in the interval 57B20;57C2, between the cv-2 and Pu loci.

Furthermore, for each mago allele, meiotic recombination was used to remove accessory mutations that might lie outside of the cv-2 to Pu interval. For each allele a number of recombinant lines were tested. In this manner, it was possible to exclude the possibility that the observed phenotypes were due to accessory mutations outside of the cv-2 to Pu interval.

Eggs were collected for egg counts and cuticle observation on 35×10 mm plates containing a molasses agar medium (Bos- well and Mahowald, 1985b). At 25 °C, eggs were collected for 12 h and the embryos were aged 2 days. At 17°C, eggs were collected for 12 h and embryos were aged for 6 days. The frequency of egg hatching was determined, and the inviable embryos were prepared for cuticle observation using the method of Van der Meer (1977). Embryos were viewed by phase-contrast microscopy.

For cytoplasmic transfer experiments, eggs were collected at 25°C during 30min intervals and, at 17°C, eggs were collected during 1–1.5 h intervals. Embryos were prepared for cytoplasmic transfer as described by Lehmann and Nüsslein- Volhard (1986). All recipient embryos were early cleavage stage embryos and donor embryos were of the stages specified in the text. Injected embryos were developed two days at 25 °C and 4 days at 17 °C.

For electron microscopy (EM) and thick-section analysis, embryos were collected on agar plates as above. Collections were made at 2–4 h intervals at 25 °C and 6–16 h intervals at 17°C. Embryos were dechorionated in 3% sodium hypo- chlorite for 1.5 min, washed 3 times with PBS (130 mM NaCl, 20 mM KP_i pH7.4) and then fixed in a 1:1 mixture of heptane: 25 % glutaraldehyde for 1-2 min Embryos were then fixed for 1.5 h in 4% glutaraldehyde, 2% formaldehyde, 2.5% DMSO, ImM CaC12, 0.1M sodium cacodylate, and 1% Acrolein. After hand dissection from the vitelline mem- branes, the embryos were rinsed 4 times for 5 min each in 0.1M sodium cacodylate and postfixed in 1% OsO₄ in 0.1M sodium cacodylate for 1.5 h at room temperature. They were then rinsed 3 times for 5 min each in distilled water, 2 times, 10 min each in 0.1M sodium acetate and stained with 0.5% uranyl acetate in 0.1M sodium acetate overnight at 0°C. The embryos were rinsed twice in distilled water for 5 min each and dehydrated in a series of solutions of increasing acetone concentrations, 10 min each followed by three 10 min washes in 100% acetone. Acetone was substituted with epon (1:1 acetone:epon/Araldite for 5h, followed by 100% epon/Araldite overnight. The embryos were then embedded in 100% epon/Araldite between two glass slides and polym- erized overnight at 60 °C. Embryos were cut out and mounted onto epon blocks for sectioning.

Longitudinal thick (0.5 mm) and thin (70 nm) sections of each embryo were taken. Both ends of the thin-sectioned embryos were inspected on a Phillips CM10 transmission EM for the presence of polar granules.

The temperature-sensitive periods for both mago¹ and mago² were determined using females heterozygous for the mutant allele and the chromosomal deletion Df(2R)F36 (O’Donnell, et al. 1989). Egg collections were done on molasses agar plates. Females heterozygous for mago¹ or mago² and Df(2R)F36 were raised at 25°C or 17°C and then shifted to 17°C and 25°C, respectively. Beginning 24 h before the shift, eggs were collected at 12 h intervals from mago¹/Df(2R)F36 females, or 8h intervals from mago²/Df(2R)F36 females. Eggs were collected continuously at these intervals for 8 days to ensure that the last eggs collected had completed oogenesis at the post-shift temperature. The eggs were aged for 2 days at 25°C or 6 days at 17 °C, and then scored for the percentage of eggs that hatched. The temperature-sensitive periods were then determined based on the effect of the temperature shift on the viability of the embryos. To determine whether the embryonic lethality and the phenotype of the inviable embryos exhibit the same temperature-critical periods, inviable embryos from mago¹/Df(2R)F36 females were collected and prepared for cuticle observation, as described above.

The three mago alleles described here are strict maternal effect mutations identified solely on the basis of the production of sterile F[progeny by hemi- and homozygous mothers (grandchildless-like phenotype). In grandchildless-like mago alleles, no apparent tem- perature dependence of the sterility phenotype is observed. As with many other loci required for germ cell formation, mutations at the mago nashi locus also cause embryonic lethality. Both the frequency of embryonic death and the phenotype of the inviable embryos are allele-specific and temperature-depen- dent. The partially permissive temperature is 25 °C and the non-permissive temperatures are below 18° and above 26°C. Here we present: (1) an analysis of the cuticular pattern defects observed in embryos produced by mago mothers maintained at 17° or at 25°C; (2) an ultrastructural analysis of the germ plasm of oocytes and embryos derived from mago females; (3) an examination of the temperature-sensitive periods of mago¹ and mago²; and (4) cytoplasmic transplantation experiments to test for the presence of functional posterior determinants in embryos collected from homozygous mago¹ females reared at 17 °C.

Inviable embryos collected from homozygous, hemizy- gous or heteroallelic mago mothers reared at 17 °C resemble embryos collected from mothers mutant for posterior group genes, such as oskar (psk) and tudor (tud) (Boswell and Mahowald, 1985b; Lehmann and Nüsslein-Volhard, 1986). In the most extreme cases, the embryos lack all ventral cuticular pattern elements of the abdomen (Fig. 1B). Approximately 84–98% of the embryos die whether the mother is homozygous, hemizygous or heteroallelic for mago¹ or mago² (see Table 1). The extreme phenotype, the complete lack of ventral abdominal cuticular pattern elements, is exhibited by approximately 90–96% of the inviable embryos (see Table 2). The remaining embryos exhibit related abdominal cuticular pattern defects that in- clude: (1) a single broad abdominal denticle belt consisting of a mirror-image duplication of similar abdominal-like hairs, or (2) embryos with two or more abdominal segments. Embryos with two or more ab- dominai segments always have an apparently normal first abdominal setal belt (Al). Therefore, at 17°C, a small fraction (<10%; see Table 2) of the inviable embryos derived from homozygous and hemizygous mago¹ or mago² females exhibit some normal abdomi- nal cuticular structures. Moreover, mago¹ and mago² display a similar, but incomplete, embryonic lethality when homozygous or hemizygous females are main- tained at 17°C. Thus, at this temperature, mago¹ and mago² are most likely hypomorphic (reduced function) mutations. Furthermore, the extreme abdominal cu- ticular defects displayed by inviable embryos derived from mago females reared at 17°C are identical to the phenotype observed in embryos derived from females homozygous or hemizygous for apparent amorphic (null) mutations in other genes of the maternal effect posterior group (Nüsslein-Volhard et al. 1987; Manseau and Schiipbach, 1989a).

However, unlike genes of the maternal effect posterior group some mutations of the mago locus appear to result in the production of altered gene products at 25 °C. As shown in Table 1, at 25 °C, approximately 60–65 % of the embryos from mago¹ and mago² homozygous females die. In contrast, at this same temperature, only 26 % and 24 % (see Table 1) of the embryos from hemizygous mago²and mago¹ females die, respectively. In all cases, the surviving F₁ progeny are phenotypically normal, but fail to form germ cells. Moreover, the frequency of embryonic lethality observed in embryos from + /mago¹ (17/713 or 2 % embryonic lethality) and +/mago² (76/736 or 10 % embryonic lethality) females is similar to that observed in control embryos from a wild-type strain, Oregon R (data not shown). Together, these results indicate that mago¹ and mago² are recessive antimorphic maternal effect mutations at 25 °C.

Also, unlike genes of the posterior group, and unlike the results at 17°C, the inviable embryos from mago females reared at 25 °C exhibit a continuum of cuticular phenotypes that most closely resemble the cuticular pattern defects observed in embryos collected from homozygous or hemizygous bicaudal females (Nüsslein- Volhard, 1977). The range of cuticular pattern abnor- malities observed is illustrated in Fig. 2. Embryos that do not fall within the above continuum display the ventral cuticular pattern defects observed in embryos collected from mago¹ and mago² homozygous and hemizygous females reared at 17°C (not illustrated but identical to the embryo in Fig. IB). The mago alleles can be distinguished, one from another, by the frequency with which they produce inviable embryos that fall into any of the phenotypic classes just described (see Table 2). For example, mago¹ homozygous fe- males produced the highest frequency of double abdomen embryos, approximately 3%. In contrast, mago² homozygous females were not observed to produce double abdomen embryos. Instead, the ma- jority of embryos from mago² homozygous females lack abdominal segments, but have normal head, thoracic and telson structures (see Table 2). Interestingly, approximately 46% of the embryos derived from mago¹/mago² mothers (reared at 25 °C) die, and of these approximately 4% exhibit the double abdomen phenotype. Furthermore, although hemizygous mago² females produce double abdomen embryos at a low frequency (˜1 %), approximately 11 % of the embryos have labrum-specific defects or labrum and abdominal defects. The labrum is the preoral lobe of the head and is the most anterior portion of the blastoderm fate map of the cuticle of the larval head (Jürgens et al. 1986). The labrum defects observed in embryos produced by mago² mothers are rarely, if ever, seen in embryos from mago¹ females. The labrum defect represents the least severe alteration of the cuticular pattern in embryos in which head structures are deleted and replaced by abdominal segments. The most severe transformation of this kind that we observe is the double abdomen phenotype observed among the progeny of mago¹ females. Thus, mago¹ is more capable of leading to a replacement of head segments by segments of a posterior origin than is mago².

The third allele, mago², is not viable when homo- zygous or hemizygous. We have examined the cuticular structures of presumed homozygous, inviable, mago³ embryos and observed no apparent cuticular defects. When embryos were collected from mago¹/mago³ or mago²/mago³ females reared at 17°C, approximately 92 and 85 % (Table 1) of the embryos die, respectively. The inviable embryos lack the ventral cuticular pattern elements of the abdomen, but exhibit normal head, thoracic and telson structures (Table 2). Approximately 53 and 44% of the embryos from mago¹/mago³ and mago²/mago³ females reared at 25°C die, respectively. At 25°C, mago¹/mago³ females produce double abdo- men embryos (˜10% of the inviable embryos), and a range of cuticular phenotypes typically observed in embryos derived from mago¹/mago¹ mothers (Table 2). However, a novel phenotypic class is also observed that consists of embryos lacking the abdominal segments, but having supernumerary posterior spiracle-like and Filzkörper structures anterior to the normal posterior structures (this class is ˜10% of the inviable embryos; see Fig. 3 and Table 2). mago²/mago³ females produce embryos that exhibit supernumerary posterior struc- tures at a similar frequency, 6%. Embryos collected from mago¹/mago³ and mago²/mago³ females can be distinguished, one from the other, because embryos from mago²/mago³ mothers do not display the double abdomen phenotype. Supernumerary posterior spiracles and Filzkörper have not been observed in embryos derived from mothers mutant in maternal effect genes of the posterior group or the spatial coordinates class, and thus, this represents a novel phenotypic class.

As stated earlier, all surviving F₁ progeny from mago homozygous, hemizygous or heteroallelic mothers are sterile, and exhibit no temperature-dependent variation of the sterility phenotype. Among mutations within the posterior group genes (cappuccino, oskar, staufen, spire, tudor, vasa and valais) that result in the complete absence of pole cells, it-has been observed that the polar granules, germ plasm-specific electron-dense cytoplas- mic structures, are reduced or absent (Boswell and Mahowald, 1985b; Lehmann and Nüsslein-Volhard, 1986; Schüpbach and Wieschaus, 1986; Manseau and Schüpbach, 1989b). In mutations at one of these loci, tud, the ability of embryos to form germ cells has been shown to be allele-specific and to correlate with the total amount of assembled polar granule material observed in the posterior pole plasm, i.e., as the amount of polar granule-like material is reduced there is a concomitant reduction in the ability to form germ cells (Boswell and Mahowald, 1985b). Thus, ultrastruc- tural analysis of the germ plasm has become a reliable assay for the integrity of the germ plasm. Therefore, to determine whether there were germ plasm defects, we examined the posterior pole plasm of embryos from mago¹ and mago² homozygous or hemizygous females at the ultrastructural level. In an analysis of 72 embryos derived from mutant mago females reared at 17° and 25 °C, none were observed to have normal polar granule morphology. In some embryos (16/72), infrequent, reduced polar granule-like material was observed in the posterior pole plasm (illustrated in Fig. 4). This material was less dense than that observed in wild-type posterior pole plasm, was associated with polysome-like structures, and frequently remained attached to mito- chondria. Thus, these less dense polar granule-like structures exhibit the typical features of polar granules, and may represent polar granule remnants. However, in the majority of embryos (56/72) no polar granule-like material was evident in the posterior pole, nor did we observe ectopic polar granules.

Rather than a specific defect in the determinative process, mutations in maternal genes required for germ cell formation and segmentation of the embryonic abdomen may instead cause: (1) a delay in the migration of cleavage-stage nuclei into the posterior pole, as observed in grandchildless of D. subobscura (Mahowald et al. 1979) and gs(1)N26 (Niki and Okada, 1981); and (2) incomplete cellularization at the blasto- derm stage, observed in valais (Schiipbach and Wie- schaus, 1986), when females are reared at 22°C. It was, therefore, possible that the defects associated with mago mutations result from a disruption of nuclear migration to the posterior pole or a failure in cellularization. Thus, to examine this possibility, embryos from mutant mago females were prepared for observation both by light and scanning electron microscopy. Neither delayed nuclear migration nor failed cellularization were ever observed (not illus- trated). Therefore, in mago mutants, neither the failure to form germ cells, nor the aberrant segmentation of the embryonic abdomen can be accounted for by delayed migration of cleavage-stage nuclei to the posterior tip or a failure to properly cellularize at the cellular blasto- derm stage.

Mutations of the mago locus are strict maternal effect mutations indicating that mago⁺ product is required during oogenesis and/or before the onset of zygotic genome transcription. It has been observed that the temperature-sensitive period (TSP) of strict maternal effect mutations is limited to oogenesis, or occurs during early embryogenesis. Two independently iso- lated mago alleles were found to be cold-sensitive mutations, and thus, could be used to measure the temperature-sensitive period of these mutants. Since mago³ is not viable in the homozygous or hemizygous state, it was not possible to determine whether this allele is a temperature-conditional mutation. Tempera- ture-sensitive mutations of the Notch and shibire loci (Poodry et al. 1973; Shellenbarger and Mohler, 1978), in D. melanogaster, have been found to express distinct TSPs for lethality and morphological defects. Mutations of mago exhibit both embryonic lethality and segmen- tation pattern abnormalities; thus, it was of interest to examine whether the two defects displayed the same or independent TSPs. Therefore, we used two different criteria to measure the TSP of mago mutants: (1) the embryonic lethality of progeny derived from hemizy- gous mago¹ or mago² females; and (2) the frequency and kind of cuticular pattern defects observed in the developing embryos laid by hemizygous mago females. Fig. 5 shows the results of reciprocal temperature-shift experiments as measured by the inviability of embryos derived from females maintained at the non-permissive temperature (17°C) and then shifted to the permissive temperature (25°) at defined intervals or vice versa (see Methods and Materials for details). Although the results for mago¹ are not described in Fig. 5, the TSPs of mago¹ and mago² are indistinguishable, by either criterion used. The TSPs for both mutations begin at approximately stage 7 and end at approximately stage 14 of oogenesis. The data clearly indicate that the TSP of mago¹ and mago² is during oogenesis, is limited to oogenetic stages 7 through 14, and that the lethality and morphological defects of mago mutations display the same TSP.

In Drosophila, cytoplasmic transfer experiments have served as a powerful tool to examine the spatial distribution and activity of particular maternal gene products (Anderson, 1987; Nüsslein-Volhard et al. 1987; Manseau and Schϼpbach, 1989a). Thus, we employed cytoplasmic transfer experiments to deter- mine: (1) whether the abdominal segmentation defects observed in embryos collected from homozygous mago females reared at 17°C could be alleviated by cytoplasm collected from wild-type embryos; and (2) if embryos collected from homozygous mago females reared at 17 °C contained a functional posterior morphogenetic center.

Wild-type donor cytoplasm can alleviate the abdomi- nal cuticular defects observed in embryos derived from homozygous mago¹ females reared at 17 °C (see Table 3). Approximately 7 % (18/265) of the uninjected embryos collected from homozygous mago¹ females reared at 17°C exhibit ≥3 abdominal segments. Therefore, the frequency with which injected embryos formed ≥3 abdominal segments was used as an assay to determine the presence of functional posterior determi- nants. Wild-type donor cytoplasm must come from 0 to 10 % egg length (EL; 0 % represents the posterior and 100% the anterior pole of the embryo) and is most effective when transplanted into the presumptive abdominal region (40–50 % EL; data not shown) of the recipient embryo. Once the donor embryos have reached the cellular blastoderm stage (˜2.2–2.8h) the donor cytoplasm is no longer able to correct the observed cuticular defects (Table 4). Thus, the site of action (40–50% EL) of the cytoplasmic components required for correcting the abdominal segmentation defects observed in mago is at a distance from the source of these components (0–10%) and, although the TSP for mago¹ is during stages 7–14 of oogenesis, the cytoplasmic components capable of alleviating the abdominal cuticular defects can act as late as ˜1.5 h in embryogenesis. Furthermore, 0–10% EL cytoplasm collected from embryos derived from homozygous mago¹ females reared at 17°C does not alleviate the abdominal segmentation defects observed in embryos derived from mago¹ homozygous females reared at 17°C, nor does it correct abdominal defects observed in embryos collected from homozygous osk¹⁶⁶ females (see Table 3). Taken together, these results suggest that functional posterior determinants are not present in the posterior pole of embryos derived from homozygous mago¹ females reared at 17°C. Therefore, these results provide direct evidence that mutations of the mago locus disrupt the ability of the posterior organizing center to function. Moreover, these results are consist- ent with results obtained when wild-type or mutant cytoplasm is transferred to embryos collected from females mutant in other posterior group maternal effect genes (Lehmann and Nüsslein-Volhard, 1986, 1987).

In this report, we have described the phenotypic characteristics associated with mutations of a newly identified maternal effect locus, mago nashi. The experimental results presented provide compelling evidence that the wild-type gene product of the mago nashi locus is vital for the establishment of longitudinal axial polarity in the Drosophila embryo. Specifically, we provide evidence suggesting that: (1) mago⁺ function is required during a discrete period of oogenesis for segmentation of the embryonic abdomen; (2) at the partially permissive temperature mutations of the locus result in the production of a novel gene function capable of altering the fates of cells along the entire anteroposterior axis; and (3) mago⁺ function is required for germ cell determination. Moreover, mago nashi is novel among maternal effect genes required both for segmentation of the embryonic abdomen and germ cell determination in two respects. First, mu- tations of the mago nashi locus can result in a deletion of anterior embryonic pattern elements and a replace- ment of the deleted elements by a mirror-image duplication of posterior abdominal and tail structures. Second, mothers heteroallelic for particular mago alleles produce embryos lacking pattern elements of the abdomen and these pattern elements are replaced by supernumerary tail structures. Thus, as discussed below, we have identified a gene that produces a crucial component of the posterior organizing center.

The cuticular pattern abnormalities associated with mutations of mago suggest that the mago⁺ product is essential in an oogenetic process necessary for segmen- tation of the embryonic abdomen. When females homozygous, heteroallelic or hemizygous for either mago¹ or mago² are reared at 17°C approximately 90 % of their progeny die during embryogenesis, and the majority (˜93%) of the inviable embryos lack normal cuticular pattern elements of the embryonic abdomen. This aberrant cuticular pattern is indistinguishable from the phenotype observed in embryos derived from females homozygous or hemizygous for apparent amorphic (null) alleles of posterior group maternal effect genes, such as osk (Lehmann and Nüsslein- Volhard, 1986). Furthermore, embryos collected from mago¹ homozygous females contain no apparent func- tional posterior determinants in the posterior pole. However, the incomplete penetrance of the embryonic lethality and the small percentage (<10%) of inviable embryos with more than two abdominal segments suggest that mago¹ and mago² are not amorphic alleles, but hypomorphic (reduced function) alleles. Thus, in the absence of a completely functional mago⁺ product (17°C) abdominal segmentation rarely occurs.

In contrast, at 25°C, there is, an increased viability of the embryos derived from mago¹ or mago² mothers, but embryos from females homozygous for these alleles exhibit a higher incidence of embryonic lethality than do embryos from mothers that are hemizygous for either allele. Although further dosage studies are necessary to determine the exact nature of the mago mutations, the genetic results described here clearly indicate that: (1) 25°C is the more permissive tempera- ture; and (2) at 25 °C mago mutations result in the acquisition of a novel gene function rather than eliminating or reducing the function of the gene. Thus, at 25 °C, mago mutations behave as recessive antimor- phic alleles (m/m>m/Df>m/+ = +/+, where m rep- resents the mutant allele and the severity of the phenotype is depicted in decreasing order; cf Muller, 1932). Moreover, at 25°C, mago¹ homozygous or hemizygous females produce inviable embryos that display a range of phenotypes representing a continuum from symmetrical double abdomens at one extreme to morphologically wild-type embryos at the other. At 25 °C, homozygous mago¹ females produce no double abdomen embryos and hemizygous mago¹ females produce them infrequently (1 % of the inviable em- bryos are double abdomen embryos). In addition, mago¹ hemizygous females produce embryos lacking labrum structures. If the continuum from double abdomen to wild-type embryos represents the graded effect of a gene product capable of inducing the replacement of head and thoracic structures by pos- terior abdominal and asegmented tail structures, then the labrum defects observed in embryos derived from hemizygous mago² females may be considered to be the least extreme expression of this replacement. The most extreme expression of this defect would then be the double abdomen embryos derived from mago¹ mothers. Most of the inviable embryos lack abdominal segmentation and, thus, represent a phenotype that does not lie within the continuum described above.

Formation of a functional posterior organizing center in the Drosophila embryo requires the products of at least nine genes (Nüsslein-Volhard et al. 1987; Manseau and Schüpbach, 1989a). Together, these genes are involved in the synthesis, assembly and proper distri- bution of the components of the posterior organizing center. Although amorphic alleles of mago are not available, it is clear that at 17°C mutations of mago disrupt two of the functions attributed to the posterior organizing center, germ cell formation and segmen- tation of the embryonic abdomen. Therefore, we propose that mago⁺ function is required for the proper assembly or distribution of the components of the posterior organizing center. The isolation and charac- terization of amorphic alleles of mago should allow us to determine whether mago⁺ function is specifically required, during oogenesis, for the formation of a functional posterior organizing center. Furthermore, epistasis studies and a determination of the distribution of maternal and zygotic gap gene products in embryos derived from mutant mago females should provide insights into the functional relationships between mago and other genes required for delineation of the anteroposterior axis.

As demonstrated by the phenotypes of the inviable embryos at 25°C (the more permissive temperature), the products of mago¹ and mago² are capable of changing the fate of cells along the anteroposterior axis of the embryo, although the mago² product is not as efficient as the mago¹ product at causing the replace- ment of anterior structures with posterior structures. Dominant mutations of the Bicaudal-D (Bic-D) locus produce double abdomen embryos by affecting the localization of both anterior and posterior determi- nants. Both the anterior and posterior pole cytoplasm of embryos collected from Bic-D¹/Bic-D² mothers contain functional posterior determinants (Lehmann and Nüsslein-Volhard, 1986). In addition, bed RNA is unstable and bed protein (the anterior morphogen) is absent (Driever and Nüsslein-Volhard, 1988a,b; Whar- ton and Struhl, 1989). Taken together, these results suggest that Bic-D mutations result in the mislocaliza- tion of posterior determinants to the anterior pole and a destabilization of the anterior determinants in the anterior pole; thus, forming an active posterior organiz- ing center at the anterior pole of the embryo. However, the germ plasm of embryos collected from Bic-D¹/Bic- D² females is localized normally and the embryos form pole cells only at the posterior end. Thus, although the specific mechanism by which mago mutations induce double abdomen embryos is not known, it is reasonable to propose that mutations at the mago locus may result either in the mislocalization of the posterior organizing center in the anterior pole or stabilization of posterior determinants in the anterior pole. With both mago¹ and mago², the majority of embryos collected from homo- zygous or hemizygous females reared at 25 °C lack normal segmentation of the embryonic abdomen. The existence of embryos that lack abdominal segmentation may be due to the lability of the mago gene product when the mago locus has been mutated, although we have no direct evidence in support of this idea. Nevertheless, this scheme does explain the range of phenotypes observed among mago embryos and is consistent with results from studies in a variety of insects suggesting that double abdomen embryos can be induced by experimental interference of normal devel- opment (Sander, 1976; Frohnhöfer et al. 1986; Kalthoff and Elbetieha, 1986) or by mutation (Bull, 1966; Nüsslein-Volhard, 1977; Mohler and Wieschaus, 1985; Suter et al. 1989; Wharton and Struhl, 1989)].

In some heteroallelic combinations (mago¹/mago³ and mago²/mago³) approximately 7 % of the inviable embryos exhibit a novel phenotype in which they lack abdominal segments, but display supernumerary pos- terior spiracles and Filzkörper structures anterior to the normal set of posterior structures. The inability of mago³ homozygotes or hemizygotes to survive, and the failure of mago³ to complement loci within the chromosomal interval containing the mago locus suggest that mago³ is a chromosomal rearrangement, although it is not cytologically visible. However, chromosomal deletions that completely remove both the mago locus and loci flanking mago produce a different range of phenotypes than do mago¹/mago³ and mago²/mago³ females. Therefore, if mago³ is a chromosomal deletion, it probably does not remove the entire locus. Whether mago³ is a small chromosomal deletion or a small inversion, it is possible that the mago³ product is modified such that components of the posterior organizing center become improperly concen- trated within the posterior end of the embryo. Thus, in these heteroallelic combinations posterior determinants are active at an ectopic site within the embryo, resulting in the production of posterior tip structures in a region that normally exhibits abdominal pattern elements.

The mago locus provides a unique opportunity to study the mechanisms involved in germ cell determi- nation and delineation of the anteroposterior axis. First, it can be considered to be a posterior group maternal effect gene required both for germ cell determination and segmentation of the embryonic abdomen. This was demonstrated both by the lack of abdominal segments in inviable embryos derived from mutant mothers reared at 17 °C and the absence of germ cells in embryos derived from such mothers. Second, at the partially permissive temperature of 25°C mago mutations produce viable embryos that lack germ cells and normal polar granules, and the in viable embryos can exhibit mirror-image duplications of posterior abdominal and telson structures. Previously, the forma- tion of double abdomen embryos has only been observed in embryos derived from mothers mutant in the spatial coordinate genes, bicaudal, Bicaudal-C and Bicaudal-D (Bull, 1966; Nüsslein-Volhard, 1977; Mohler and Wieschaus, 1985). Disruption of germ cell formation, however, is never observed in embryos from mothers mutant for spatial coordinates genes. This indicates that mago nashi exhibits features of both the posterior group and spatial coordinate maternal effect genes. Thus, the mago locus is unique in that embryos derived from mutant mothers lacking a functional germ plasm can form posterior abdominal segments and form posterior abdominal structures in place of anterior pattern elements of the head and thorax. Therefore, molecular and further genetic analysis of the mago locus and its interactions with other genes involved in establishing anteroposterior polarity should provide new insights into the mechanisms underlying early pattern formation.

We thank Phillip Newmark and Drs Susan Dutcher, George Golumbeski and David Prescott for critical reading of the manuscript. We are grateful to Virginia Fonte for technical assistance with transmission electron microscopy. This work was aided by grant no. NP-52138 from the American Cancer Society. R.E.B. is also a recipient of an award from the PEW Scholars program.